Quaternary Science Reviews 19 (2000) 189}211

Hydrological changes in the African tropics since the Last Glacial Maximum Franc7 oise Gasse CEREGE, B.P. 80, 13545, Aix-en-Provence Cedex 04, France

Abstract

Paleohydrological data from the African tropics and subtropics, including lake, groundwater and speleothem records, are reviewed to show how environments and climates from both hemispheres are inter-related. Although orbitally induced changes in the monsoon strength account for a large part of long-term climatic changes in tropical Africa, the Late Pleistocene}Holocene hydrological #uctuations rather appear to have been a series of abrupt events that re#ect complex interactions between orbital forcing, atmosphere, ocean and land surface conditions. During the Last Glacial Maximum (23}18 ka BP), most records indicate that generally dry conditions have prevailed in both hemispheres, associated with lower tropical land- and sea-surface temperatures. This agrees with simulations using coupled ocean}atmosphere models, which predict cooling and reduced summer precipitation in tropical Africa; the global hydrological cycle was weaker than today when the extent of large polar ice-sheets and sea-ice was a prominent forcing factor of the Earth's climate. Glacial-interglacial climatic changes started early: a "rst wetting/warming phase at ca. 17}16 ka BP took place during a period of rapid temperature increase in Antarctica. Next, two drastic arid-humid transitions in equatorial and northern Africa occurred around 15}14.5 ka BP and 11.5}11 ka BP. Both are thought to match the major Greenland warming events, in concert with the switching of the oceanic thermohaline circulation to modern mode. However, part of the climatic signal after 15 ka BP also seems related to the Antarctica climate. During the Holocene, Africa has also experienced rapid hydrological #uctuations of dramatic magnitude compared to the climatic changes at high latitudes. In particular, major dry spells occurred around 8.4}8 ka and 4.2}4 ka BP in the northern monsoon domain. Comparison with other parts of the world indicates that these events have a worldwide distribution but di!erent regional expressions. In the absence of large polar ice sheets, changes in the continental hydrological cycles in the tropics may have a signi"cant impact on the global climate system. Climate information gathered here allows to identify geographical and methodological gaps, and raise some scienti"c questions that remain to be solved to better understand how the tropics respond to changes in major climate-forcing factors, and how they in#uence climate globally. ( 1999 Elsevier Science Ltd. All rights reserved.

1. Introduction hundreds of metres above their present level. Conversely, from about 20,000 to 15,000 years ago, the largest lake in The hydrological #uctuations that took place in Africa Africa, Lake Victoria, was desiccated (Talbot and Living- during the Late Quaternary represent a impressive mani- stone, 1989; Johnson et al., 1996), as were many others. festation of continental climate change. In recent times, The distribution and density of the human population in exceptional rainfall events over East Africa, associated Africa is regulated primarily by the availability of water with El Nin o-Southern Oscillation (ENSO) years to drive biological and human activities. It is thus of (Nicholson, 1996), have generated catastrophic #oods. relevance to study past hydrological records to under- During the Sahel drought in the 1970s and 1980s, mean stand how regional environmental changes are related to annual rainfall declined by 30% (Hulme, 1992); cattle changes in the Earth's climate system. were decimated and people migrated southward. How- This paper focuses on the African tropics and sub- ever, these recent events do not rival those of past mil- tropics within PAGES-PEP III Time Stream II (the last lenia. From about 10,000 to 4000 years ago, Neolithic 250,000 yr at time resolution of 10}10 yr; PAGES, civilizations #ourished in a wet and green Sahara, while 1997). It aims to use paleohydrological proxy data to in northern East Africa closed lakes extended tens or document how environments and climates from both hemispheres are inter-related. Regional paleohydrologi- cal records, derived from lake, groundwater, and spele- E-mail address: [email protected] (F. Gasse) othem archives, are compared. Some pollen data

0277-3791/99/$- see front matter ( 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 7 7 - 3 7 9 1 ( 9 9 ) 0 0 0 6 1 - X 190 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 associated with lake studies are also considered. Most 1996). The East}West climate asymmetry is linked to records do not provide the resolution and continuity that topographical features (e.g., the eastern escarpment of would be required for detailed correlations with high southern Africa), and to sea surface conditions (e.g., the latitude time-series, especially with the polar ice core cold oceanic Benguela, and the warm Algulhas currents records. Nevertheless, the paleohydrological data assem- #owing along the western and eastern coasts of southern bled here can be useful for model-data comparisons, Africa, respectively). yielding hypotheses on the linkages between high and The major hydrological features of Africa result from low-latitude climates. The paper is structured around the the general climate, topography and geological patterns. following questions: They are (Fig. 3): (i) large internal drainage basins, e.g., the Lake Chad basin, and enormous groundwater reser- E How did African climates in both hemispheres re- voirs, e.g., the geological formations of the `Continental spond to orbital forcing versus other glacial boundary Intercalairea in the Sahara; (ii) large exoreic rivers, the conditions during the last glacial maximum? Congo, the Nile and the Niger Rivers and associated E How were the successive steps of the last glacial}inter- wetlands, which have their headwaters in the humid glacial transition in Africa linked with deglacial events equatorial belt; (iii) great East African Rift lakes which at high latitudes? play a signi"cant role in regional climate through water E How did abrupt Holocene changes correlate with cli- recycling; (iv) other permanent waterbodies concen- mate changes elsewhere in the world, and how did they trated in, or depending on the equatorial zone (e.g., Lake a!ect the availability of water resources? Chad through the Chari River), or on highlands (e.g., the Ethiopian lakes Ziway-Shala and AbheH ). African lake records are mainly based on geomorphol- 2. Major climatic and hydrological features, and ogy, sedimentology, stable isotope contents of primary paleohydrological archives lacustrine carbonates and organic matter, and biological remains (e.g., diatoms, ostracods). Several authors give African climates depend on low altitude pressure and site locations and syntheses at a continental scale (e.g., winds over the continent, which are the surface expres- Street-Perrott and Perrott, 1993; Jolly et al., 1998). sion of the upper air circulation. Climates exhibit Readers may refer to the compilation on `lake statusa a broadly zonal pattern with varying seasonal distribu- (high, intermediate or low water level) by Street-Perrott tion of precipitation (Figs. 1 and 2). The northern and et al. (1989), and the Global-Lake Level Database avail- southern ends of the continent are a!ected by the equa- able at the World Climate Data Center-A (WDC-A, torward displacement of the mid-latitude westerlies Boulder, CO), but the original papers used for their during winter. These temperate regions experience Medi- construction should also be consulted, since the `lake terranean summer-dry climates, receiving most precipita- statusa is an interpretation which sometimes deviates tion during winter from westerly cyclonic disturbances. from that in the original works. They are both #anked by subtropical deserts, the Sahara Palaeoclimatic interpretation of African lake records north of the Equator, and the Namib coastal desert in often remains limited for several reasons. First, few works southwestern Africa, which are dominated by subtropical discuss thoroughly the reliability of their radiocarbon anticyclones throughout the year. These arid zones are chronologies, although C age distortions are numerous separated by a wide belt of tropical climates. The tropical in African lakes (Gasse and Fontes, 1992). Second, high climate is governed by the seasonal migration of the time-resolution lake records are rare. Finally, a full inter- Intertropical Convergence Zone (ITCZ; the meteorologi- pretation of the lake water, salt or isotope budgets at any cal equator) in response to changes in the location of given time should be based on the analysis of the re- maximum solar heating. This results in northern and sponse timing and processes of the hydrological system southern belts of monsoonal climates with summer rains (lake and catchment area) to change in the precipitation and winter drought, bracketing a humid equatorial zone minus evaporation balance (P}E). Estimates of past characterized by a double rainfall maximum. The zonal P and E from lake records, based on biological or temperature and moisture patterns are altered by high- sedimentological calibration functions and hydrological lands which act as water towers for the surrounding modelling, is possible as exempli"ed below. As yet, few lowlands. The zonal (`Walkera) circulation along the approaches of this type have been conducted in Africa. equator, caused by East}West di!erences in surface and Reliable palaeoclimatic interpretation is particularly dif- tropospheric temperatures, also greatly in#uences tropi- "cult for groundwater-fed lakes such as those which cal climates. Rainfall #uctuations in many areas of the prevailed in the Sahara and the Sahel. Groundwater-fed continent are statistically linked to the ENSO, but these lakes can be sensitive indicators, provided that the re- may be more directly a response to sea-surface temper- charge area is restricted, the #ow patterns are known, ature (SST) #uctuations in the Indian and Atlantic and the response time to precipitation change is short Oceans which occur in the context of ENSO (Nicholson, (Gasse et al., 1990). F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 191

Fig. 1. Schematic of the general patterns of winds and pressure over Africa. Dotted lines indicate the Intertropical Convergence Zone (ITCZ), dashed lines indicate the Congo Air Boundary. From Nicholson (1996). 192 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

(Stuiver and Reimer, 1993) back to 18,000 C yr BP, and the equation proposed by Bard et al. (1997) for older ages. In the text below, ages are expressed in ka, which means 10 cal. yr BP. Sites have been numbered, S1, S2,2, and are shown in Fig. 3.

3. The last glacial period

The palaeohydrological data in this section were gathered to identify factors, other than orbital forcing, that have controlled the #ux of energy and water vapour to the African continent during the last glacial maximum (LGM; ca. 23}18 ka). These data complement pollen and lake organic matter studies which have shown that decreased CO atmospheric concentration induced a lowering of mountain forests and in#uenced the carbon cycle in mountain lakes in the tropics during glacial times (e.g., Street-Perrott et al., 1997; Jolly and Haxeltine, 1997).

3.1. Evidence for cooler and drier conditions during the LGM Fig. 2. African precipitation regimes. Annual harmonics of precipita- tion, from HsuK and Wallace (1976), adapted by deMenocal and Rind A long (90 m) and continuous (200 kyrs) evaporite-rich (1996). Rainfall time series from individual climatological stations were depositional sequence was obtained from site S1, a me- averaged into monthly means and then subjected to harmonic analysis which de"ned the phase and amplitude of the annual precipitation teoric crater lake, the Pretoria Saltpan (Partridge et al., cycle. The vector length indicates normalized amplitude. Vector direc- 1997). Lake sediment texture and chemical composition tion indicates month of maximum precipitation. Southward-pointing were quantitatively related to precipitation in the catch- vectors indicate January 1 rainfall maximum; westward vectors indicate ment area. The resulting south monsoon precipitation April 1 maxima (HsuK and Wallace, 1976). time series is characterized by periodic (23,000 yr) varia- tions, attributed to the orbital precession cycle (Fig. 4). Nevertheless, the 23,000-yr cyclicity is blurred after Groundwater records may help interpret ground- ca. 50 ka. The precipitation time series shows a negative water-fed palaeolakes. In addition, dated groundwater shift starting at about 30 ka, and ending at the LGM may provide information about palaeoclimatic condi- when estimated precipitation values are about 15}20% tions when and where the aquifers recharged, through lower than today. their major and trace element content, stable isotope The stable isotope record of a stalagmite from Bot- composition and noble gas content (Stute et al., in swana (S2) covers the interval 51}21 ka (Holmgren et al., PAGES, 1997). Indeed, in con"ned (closed) aquifer 1995). The interval 51}43 ka was warm and wet. Dry systems, much of the water is fossil water in"ltrated periods at 46, 43, 26, 24 and 22 ka have been associated during past humid periods. Groundwater archives act as (Holmgren et al., 1995) with the Heinrich cooling events a low-pass "lter and thus only provide low-resolution H5 to H2 (Bond et al., 1992). After 27 ka, gradually drier time-series; nevertheless, they lend themselves to quantit- conditions "nally halted speleothem growth at ca. 21 ka. ative reconstructions of temperature, and yield informa- Glacial cooling is estimated at about 2}33C. tion on moisture transport patterns (Stute and Talma, A noble gas temperature and isotope (dO) record 1998). Speleothems provide information about past from palaeogroundwaters at Stampriet (S3, Fig. 5) covers P!E balance, vegetation cover in the catchment, the past 40 ka (Stute and Talma, 1998). At the LGM, it palaeotemperatures and isotopic composition of precipi- shows a glacial cooling as much as 5.33C lower than tation, through their texture, stable isotope content of today, similar to that observed at the Uitenhage aquifer calcite crystals, and their #uid inclusions (Lauritzen, in (Fig. 3) in the southern Mediterranean climatic zone. At PAGES, 1997). Stampriet, the present-day rainfall originating from the The chronological framework is mainly based on Indian Ocean exhibits a marked isotopic depletion due radiocarbon ages (C yr BP), and some Th/U to the continental e!ect. The dO values higher than ages. Radiocarbon ages were converted into calendar today during glacial times, associated with low temper- estimates (cal. yr BP) using the CALIB 3.0 program atures, suggest a shift in moisture transport from the F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 193

Fig. 3. Major lake and river systems in Africa, and location map of sites (S) cited in text. Sites I are from the southern monsoonal domain, II from equatorial-subequatorial Africa, III from the Sahelian belt and the southern Sahara, and IV from the winter rain domain of the northern Sahara (Fig. 2). 1: Pretoria Saltpan. 2: Lobatse. 3: Stampriet. 4: Marine core GeoBio1023. 5: Marine core MD79257. 6: L. Tritrivakely. 7: L. Malawi. 8: L. Tanganyika. 9: L. Kivu. 10: L. Victoria. 11: L. Albert. 12: L. Magadi. 13: L. Naivasha. 14: L. Nakuru-Elmenteita. 15: L. Turkana. 16: Ziway-Shala lake system. 17: L. AbheH . 18: Jebel Marra. 19. L. Barombi Mbo. 20: Marine core CH22KW31. 21: Lake Bosumtwi. 22: L. Chad. 23: Bahr-el-Ghazal.. 24: Nigeria, recharge zone of the Middle aquifer of the Chad Formation.. 25: L. Bal, Kajemarum Oasis. 26: Bougdouma. 27: Termit. 28: Bilma. 29: Tin Oua!adene. 30: Adrar Bous. 31: Recharge zone of Northern Niger, and 32: Northern Mali. 33: Western Fezzan. 34: Recharge zone of the `Continental Intercalairea aquifer. 35: Sebkha Mellala. 36: Hassi el Mejnah. 194 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

Fig. 4. Orbital precession as a dominant factor regulating late Pleistocene monsoonal precipitation in long-term records. Redrawn from Partridge et al. (1997). Comparison of: (c) variations in South monsoon precipitation reconstructed from a 200 kyr sedimentary record from the Pretoria Saltpan, South Africa (Partridge et al., 1997), with: (b) a monsoonal precipitation index at 203N based on fossil faunal assemblage variations in deep-sea sediment core RC24-07 (203N; McIntyre et al., 1989); and (a) changes in summer solar radiation in the northern and southern subtropics (after Berger, 1978).

Indian Ocean (as it is today), to the Atlantic Ocean: the No lake record is available for southwest Africa, but shorter western path for rainwater from the Atlantic a marine pollen record from core GeoBio1023 (site S4) Ocean reduced the continental e!ect (Stute and Talma, registers climate changes in the continent between 213 1998). This scenario implies that the westerly winds on and 133S, west of 243W (northern Namib desert, the western seaboard extended northward, at least to Angola-northern Namibian highlands, northwestern 253S during the Glacial period, in agreement with the Kalahari), since 21 cal. ka BP (Ning Shi et al., 1998). The climatic model of Cockcroft et al. (1987). record shows cold and dry LGM conditions. Alkenone F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 195

sional (seasonal?) rains may have occurred. The lake was at least seasonnally dry from about 22 to 17.5 ka. A highstand at Lake Malawi (S7) from about 32.5 to 11.5 ka suggested by sedimentological studies (Finney et al., 1996; Fig. 7) appears as an exception within the southern tropics. In equatorial East Africa, Lake Tanganyika (S8) was at least 300 m lower than today at ca. 21 ka (Fig. 7; Gasse et al., 1989). A water and energy balance model, established for the modern Lake Tanganyika system, was used to estimate past mean annual precipitation and evaporation at the LGM (Bergonzini et al., 1997). Input values used to simulate the LGM situation are insola- tion, past lake and catchment areas, pollen-inferred tem- perature (!4.2$23C; ChalieH , 1995), and albedo as estimated from vegetation changes in the region (Vin- cens, 1991; Bonne"lle et al., 1992). In percent of modern means, the simulation yields decreased evaporation from the lake [!5% (!13% to #3%)] and land [!8% (!19/#5)] bodies, as well as precipitation [!14% (!24/!2)]. Decreases in P, E and P}E are substan- tially ampli"ed when including empirical changes in at- mospheric transmission coe$cient (Tyson et al., 1997). Sensitivity runs suggest that even large changes in cloud cover and air humidity should not modify these trends. Lake Victoria (S10) was desiccated or nearly so at 20.5 ka (Talbot and Livingstone, 1989). The level of Lake Albert (S11) declined markedly after ca. 22 ka, and was at least 46 m lower than today between 20.2 and 18 ka (Beuning et al., 1997; Fig. 7). The seasonal discharge of the Blue Nile River was considerably reduced (Said, 1993). In subequatorial West Africa, generally dry LGM conditions are indicated by pollen records from Lake Fig. 5. Groundwater archives as paleoclimatic indicators in Africa. Barombi Mbo (S19; Maley and Brenac, 1998) and Lake Noble gas temperature and dO versus radiocarbon age for the Stam- Bosumtwi (S21; Maley, 1991), but the interval 23}18 ka is priet aquifer (Namibia). After Stute and Talma (1998). complex. At Lake Bosumtwi, a dO and dN record in lacustrine organic matter spanning the last 32,000 ka shows that the setting of a drier climate, starting at ca. 29 ka, was not linear: dry excursions centred at 25.9, 21.9, paleothermometry also indicates sea-surface temperature and 17.5 ka were followed by relatively humid phases, (SST) 3}43C lower than today at neighbouring sites notably that from 21 to 18.5 ka (Talbot and Johannessen, (Schneider et al., 1995). At a similar latitude on the Indian 1992; Fig. 8). Ocean side, SST alkenone-temperature from core In the Sahel, the recharge of the con"ned `Middle MD79257 (S5) shows that the glacial cooling (about aquifer of the Chad Formationa in northern Nigeria !2.53C) culminated around 20}19.5 ka (Bard et al., (S24), occurred between ca. 28 and 23.2 ka, and then 1997; Fig. 6). In the Madagascar highlands, a small crater stopped (Edmunds et al., 1999). Noble gas studies suggest lake, Tritrivakely (S6), has provided a multi-proxy record that the recharge temperature was at least 63C lower than spanning the last 40 ka (Sifeddine et al., 1995; Williamson at present. The end of this cool and wet recharge period et al., 1998; Gasse and Van Campo, 1998; Fig. 6). Al- was attributed to the onset of LGM arid conditions, as though the glacial period has been punctuated by several documented by lake records in the region (e.g., Servant short-term warm-dry intervals, the terrestrial pollen and Servant-Vildary, 1980). The extension of Lake Chad record shows generally low temperature conditions until (S22) at the LGM was estimated at 7% of its modern ca. 17 ka. In the lake, diatom and sedimentary records area (Adams and Tetzla!, 1985). Eastward, lakes depend- show a positive P}E balance from about 38 to 32 ka, ing on the Ethiopian highlands (S16}S17) were low dur- followed by a step-wise desiccation trend. The LGM was ing the LGM (Gasse, 1977; Gasse and Street, 1978; a period of mean annual water de"cit, although occa- Gillespie et al., 1983). Lake AbheH (S17; Fig. 7) experienced 196 .Gasse F. / utraySineRe Science Quaternary v es1 20)189 (2000) 19 iews } 211

Fig. 6. Summary results from the Lake Tritrivakely core record (Madagascar highlands). Comparison with reconstructed SST in core MD79257 in the Mozambique Channel. (a) Alkenone paleotemperature in core MD79257 (after Bard et al., 1997). (b) Mountain forest taxa in the terrestrial pollen spectrum, re#ecting changes in both atmospheric CO concentration and temperature. Abrupt #uctuations are attributed to temperature changes. (c): arboreal pollen taxa from middle-altitude forests. (d) Aquatic plant pollen. Percentage is calculated on the terrestrial pollen sum. (e): Diatom diagram of taxa from permanent, slightly alkaline, mesotrophic water. (f) Diatom diagram of aerophilous, acidobiontic taxa re#ecting subaerial, ombrotrophic conditions. Poor preservation and the absence of diatom valves indicate ephemeral conditions or seasonal/pluriannual desiccation phases. (g) S-ratio, a sediment magnetic parameter re#ecting changes from high run-o! and reducing bottom conditions (high S-ratio values) to low run-o! and oxic environments (low S-ratio values). (b) to (f): after Gasse and Van Campo (1998); (g) after Williamson et al. (1998). F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 197

Fig. 7. The amplitude of water-level #uctuations in rift lakes from East Africa. Examples of: (a) Lake Malawi (after Finney et al., 1996 and Johnson, 1996). (b) Lake Tanganyika, Planktonic diatom percentage in core MPUXII (after Gasse et al., 1989). (c) Lake Albert. Lake level curve based on data from Talbot and Livingstone (1989) and Beuning et al. (1997). (d) Lake Turkana (after Johnson, 1996). (e) Lake AbheH (after Gasse and Street, 1978).

a step-wise shrinking of more than 160 m between about lake (S18) experienced its minimum level at ca. 20.2 ka 27 and 20 ka when the lake dried up. The White Nile (Williams et al., 1980). River channel was partly blocked by sand dunes (Said, In the northern Sahara, the chronology of the few 1993). In the southern Sahara, the Jebel Marra crater records sometimes assigned to the LGM on the basis of 198 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

Fig. 8. Left: summary of some African lake and pollen records (20}8 ka). (a) Lake Tritrivakely. Frequency of woody mountain pollen taxa (as for Fig. 6b). (b) Lake Tanganyika, core MPU XII (southern basin). Percentage of planktonic diatom taxa (as for Fig. 7b) and pollen-inferred di!erence (᭿) from modern temperature and precipitation (after ChalieH , 1995, and F. ChalieH , pers. comm., 1999). (c) Marine core GeoB1023. Percentage pollen diagram of desert/semi-desert plants showing arid intervals at 14.4}12.5 and 10.9}9.3 ka (adapted from Ning Shi et al., 1998). (d) Lake Bosumtwi. Changes in dC-values of total organic matter (after Talbot and Johannessen, 1992). Positive excursions are interpreted as the result of arid climatic intervals which favoured a dominance of C4 plants in the crater and caused evaporatively driven changes in the chemical and isotopic composition of surface water dissolved inorganic carbon pool. (e and f) Changes in O content of authigenic carbonates in waterbodies from the southern (e: Bougdouma) and northern (f: Sebkha Mellala) margins of the Sahara, indicating #uctuations in the ratio of groundwater in#ux to evaporation. Negative excursions re#ect wet pulses. High values re#ect arid intervals (after Gasse et al., 1990). Right: GRIP (Greenland), Vostok and Byrd (Antarctica) isotopic records (after Johnsen et al., 1972, 1992; Jouzel et al., 1987, 1992), and CH (GRIP and Byrd) records on a common time scale (Blunier et al., 1998). Vertical dashed bands show the major warming/wetting phases in the African tropics during the last deglaciation. YD: Younger Dryas; ACR: Antarctica Cold Reversal. F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 199

C ages (e.g., Street-Perrott et al., 1989; Jolly et al., 1998) tion. It is now established that tropical SST fell by several has to be reassessed but U/Th dating (Causse et al., 1988). degrees C at the LGM (Guilderson et al., 1994; Schneider The aquifer of the `Continental Intercalairea (S34) ex- et al., 1995; Beck et al., 1997; Bard et al., 1997), contrary perienced a period of intense recharge from ca. 45 to to CLIMAP (1981) estimates. Lower tropical SST in- 23.5 ka when noble gas-inferred temperature was 53C duced a decrease in evaporation at low latitudes, the lower than the modern mean annual temperature (Guen- major source of water vapour, that further intensi"ed douz et al., 1998). Water supplying the aquifer was de- cooling through a decrease in the atmospheric concentra- pleted in heavy isotopes compared with late Holocene tion of this greenhouse gas. Recently, a coupled palaeowater and modern mean precipitation. From ocean}atmosphere climate model simulates a cooling and Morocco to the Western Desert of Egypt, an overall a decrease in summer precipitation in most of the tropics, west-to-east decrease in heavy isotope content of Late although a slight increase in winter rainfall occurs in some Pleistocene fossil groundwaters is observed, suggesting regions of southern Africa (Ganopolski et al., 1998a); the that the relative in#uence of the mid-latitude westerlies tropical cooling is due to increased oceanic heat transport was reinforced on the Sahara (Sonntag et al., 1978; Sultan out of the tropics induced by enhanced meridional temper- et al., 1997) during the glacial period, in symmetry with ature gradient and to stronger northern hemisphere trade the Namibian desert. winds, and to the reduction of atmospheric water vapour To sum up the above data, much of the continent and CO concentrations. Bush and Philander (1998) pres- experienced cooler conditions than today during glacial ent a LGM simulation which shows that a large part of the times, and a marked decrease in P or in P}E at the LGM. LGM temperatures lowering is due to a decrease in water vapour atmospheric concentration; the global hydrological cycle is weaker: both evaporation and precipitation are 3.2. African climates, precession cycles, and glacial reduced by about 10%. Model simulations thus provide boundary conditions. some explanation for the geological observations sum- marized here, although, in all models, precipitation simu- The main points arising from these observations are as lations are less reliable than temperature predictions follows. (Ganopolski et al., 1998a). (3) Spatial and temporal variations are observed. (1) Due to the geometry of orbital precession, changes The apparent exception of the Lake Malawi basin, in summer insolation are in antiphase between hemi- believed wet at the LGM may be due to the complexity spheres (Berger, 1978: Fig. 4). Increased summer insola- f the regional topography and rainfall pattern. Some tion should result in reinforced monsoon circulation by records (e.g., S19, S21) show that the interval 23}18 ka enhancing the ocean-land pressure contrast which draws has been complex, with relatively wet periods e.g. around the monsoon winds inland (Kutzbach et al., 1993; Kut- 19 ka. zbach and Street-Perrott, 1985). Orbital forcing predicts dry climates in the northern tropics, and enhanced monsoon rainfall in the southern tropics during the 4. The last glacial}interglacial transition LGM. In the northern hemisphere, long deep-sea records (e.g., McIntyre et al., 1989; Clemens and Prell, 1991) The question of interhemispheric symmetry or asym- have demonstrated that the Pleistocene monsoonal metry of major climate changes during the last deglaci- climates have been paced by Milankovitch cycles ation is currently a matter of debate (e.g., Sowers (Fig. 4); low lake-levels and low aquifer recharge and Bender, 1995; Lowell et al., 1995; Bard et al., 1997; during the LGM are consistent with orbital forcing. In Ariztegui et al., 1997; Steig et al., 1998; Blunier et al., the southern subtropics, the Pretoria Saltpan time series 1998). Discussions are mainly based on polar ice core of summer precipitation (Partridge et al., 1997; Fig. 4) records (Fig. 8). While the Greenland isotopic records provides a test of the role of orbital forcing on global (Johnsen et al., 1992; Blunier et al., 1998) evidence climate from 200 to 50 ka, although results should two major abrupt warmings at the onset of the BoK lling be con"rmed by other records supported by a better and Preboreal periods (14.6 and 11.6 cal. kyr BP, respec- chronology. tively), the temperature increase in Antarctica began as (2) Nevertheless, the fact that the Pretoria Saltpan soon as 21 cal. kyr BP at Byrd and 18.5 at Vostok climate became drier after 30 ka does not "t with the (Johnsen et al., 1972, 1992; Jouzel et al., 1987; Blunier et Milankovitch theory. Dry LGM conditions observed at al., 1998). Paleoclimatic changes in Africa in the sites S1, S2, S4 and S6 in the southern tropics cannot be time interval 20}8 ka (Figs. 7 and 8) appear compatible accounted for by changes in the Earth's orbital para- with both an early ending of full glacial condi- meters. Hydrological records appear to show that the tions, as observed at Vostok or Byrd, and further major tropical African climate has rather responded to changes steps toward full interglacial climates, as observed in in sea-surface conditions and ocean}atmosphere interac- Greenland. 200 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

4.1. Evidence for the step-wise onset of postglacial isotope record in lacustrine organic matter (Fig. 8) regis- climate. ters relatively wet periods between 21.5}18.5 and 17}15 ka followed by very dry intervals centred around At site S1 (Pretoria Saltpan), a trend toward higher 17.5, 14.0 and 12 ka, and a dramatic wetting event at monsoon precipitation is observed from about 18 ka, 11.5}10.5 ka (Talbot and Johannessen, 1992). The latter although precipitation remained low until the mid-Holo- is coincident with a major lake-level rise (Talbot et al., cene (Partridge et al., 1997; Fig. 4). Several terrestrial 1984) when the closed forest settled (Maley, 1991). At records in South Africa indicate that increased wetness is Lake Barombi Mbo (S19), changes in the Cyperaceae associated with warming which started at 17}16 ka (Par- pollen frequencies suggest that the lake level, inter- tridge, 1997); dry conditions then established during the mediate around 19, 16 and 15}14 ka, later dropped early Holocene (Partridge, 1997; Scott, 1993). At Stam- drastically; the major rise occurred at 11.5}11 ka (Maley priet (S3), deglacial warming started at 17.5}17 ka, and and Brenac, 1998). was about 60% achieved at ca. 16 ka (Fig. 5; Stute and North of 53N, the prominent postglacial hydrological Talma, 1998). The marine pollen record at site S4 (Ning change was the #ooding of a multitude of depressions at Shi et al., 1998) shows a marked temperature and humid- the onset of the Holocene. In the Sahelian belt, it has long ity increase from ca. 17.5 to 16.5 ka, and very dry periods been established that the monsoon reactivation occurred at 14.4}12.5 ka and at 10.9}9.3 ka (Fig. 8). At Tritrivakely in two steps, at ca. 15}14.5 and 11.5}11 ka, separated by (S6; Figs. 6 and 8), the interval 22}7 ka shows a very low a return to drier conditions coincident with the YD (e.g., sedimentation rate and some short-term desiccation epi- Street-Perrott and Perrott, 1990; Gasse et al., 1990; Gasse sodes. However, detailed pollen and diatom studies indi- and Van Campo, 1994; Fig. 8). The former step only cate that warm conditions and a water body favourable induced a water-level rise of moderate amplitude (Fig. 9). to aquatic life re-established in two steps, around 17.5}17 Earlier moisture increases are, however, indicated by and 15 ka, respectively (Gasse and Van Campo, 1998). isotope and sediment studies of a marine core (S20) from The Lake Malawi record (S7) suggests high lake level the Niger River delta: "rst runo! increases in the Niger until ca. 11.5 ka, when the lake level declined rapidly basin occurred prior to 15.8 ka (Pastouret et al., 1978; (Finney et al., 1996; Fig. 7). Durand, 1993). In the southern Sahara, the Jebel Marra In the equatorial East African Rift lakes, the major crater lake (S18) was 7.5 m higher than the lowest LGM event is the lake basins re"lling around 15 cal. kyr BP, level at 16.5 ka (Williams et al., 1980). In northern Niger but an earlier wetting is apparent, although poorly dated. (S31) and northern Mali (S32), dated groundwaters sug- In Lake Tanganyika (S8; Figs. 7 and 8), a generally gest that the post-glacial recharge of the aquifers began positive water budget was restored from ca. 21 ka; a "rst around 16.5 ka (Fontes et al., 1993); shallow lakes and positive oscillation may have occurred around 17 ka; swamps developed in the depressions around 15 ka (e.g. deep conditions and perennial strati"cation were estab- at S30; Gasse and Fontes, 1992). In Libyan Sahara (S33), lished by 15 ka (Gasse et al., 1989). From pollen-based U/Th datings of travertines indicate a wet episode in the reconstructions in the Southern Tanganyika basin, pre- Tadrart Acacus massif from 15.6 to 9.7 ka (Carrara et al., cipitation and temperature steeply increased from ca. 1998). At the northern margin of the Sahara, the Sebkha 18}17 to 15}14.5 ka (ChalieH , 1995; F. ChalieH , pers. com. Mellala record (S35; Fig. 8) shows a "rst increase in 1999; Fig. 8). At Lake Victoria (S10), two major late precipitation in the M'zab heights or in the Saharan Pleistocene lowstands were registered by palaeosols, and Atlas mountains centred at 16.4 ka; further short-term were separated by a minor transgression: the earlier wet pulses occurred at ca. 15 and 14.0 ka, before an arid and more intense dry interval was between ca. 20.5 and period (12.3}10.3 ka) that preceded the onset of Holocene 17.9 ka, and the second ended at ca. 15.3 ka (Talbot and permanent lacustrine conditions (Gasse et al., 1990). Livingstone, 1989), when the basin re"lled rapidly (Joh- The successive steps in the re-establishment of rainfall nson et al., 1996). This record "ts well with the hydrologi- in North Africa show good consistencies with the major cal history of Lake Albert (S11; Fig. 7): a brief and limited abrupt events of Indian monsoon intensi"cation re- transgression is also evidenced by lake sediments corded in core 74 KL from the Arabian Sea (Sirocko bracketed between two palaeosols that developed bet- et al., 1996), and dated at 16.0, 14.5 and 11.45 ka. ween ca. 20.7 and 15.3 ka; shortly after 15.3 ka, a major transgression led to open-water conditions (Beuning 4.2. Implications for north}south climatic connections et al., 1997). A dry episode "tting the cool Younger Dryas interval (YD) has been identi"ed in several equatorial Summing-up the above observations, the last deglacial lakes, e.g. Lake Kivu (S9; Haberyan and Hecky, 1987) period in Africa appears to have been a series of abrupt and Lake Magadi (S12; Roberts et al., 1993). transitions from arid to humid, or humid to arid condi- In subequatorial West Africa, a major arid}humid tions. Major events occurred around 17}16, 15}14.5, and transition starting around 11.5 ka follows several cli- 11.5}11 ka, but their expression and relative strength matic oscillations. At Lake Bosumtwi (S21), the stable varies with latitude (Figs. 7 and 8). F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 201

Fig. 9. Changes in lake level over the past 13 ka in the northern monsoon domain of Africa. After Gasse and Van Campo (1994). From East to West: (a) Lake AbheH (from Gasse, 1977); (b) Lake Ziway-Shala (from Gillespie et al., 1983); (c): Bahr-el-Ghazal (from Servant and Servant-Vildary, 1980); (d) Lake Bosumtwi (from Talbot et al., 1984). Dashed lines underlined weak monsoon episodes. 202 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

The 17}16 ka event is an arid}humid transition of low Extremely wet conditions established in equatorial amplitude in both hemispheres. Associated warming is and northern Africa at 11.5}11 ka. At the same time, documented in southern Africa (S3, S4, S6) and in the African climates south of 103S turned drier when the southern Lake Tanganyika basin (S8). It appears coinci- warming trend stopped in Antarctica (Fig. 8). Only after dent with the largest part of the temperature increase at glacial boundary conditions decayed, did Southern Afri- Byrd and Vostok in Antarctica, with the starting of the ca become sensitive to orbital forcing, in agreement with increase in global atmospheric methane concentration climate model sensitivity experiments at 10 ka (e.g., de- (Fig. 8), and with rapid climate changes recorded at many Menocal and Rind, 1996). other sites in the southern hemisphere. These include: Four conclusions arise from these observations. SST increase in the Southern Ocean (Pichon et al., 1992); rapid warming around 16.5 ka which induced a (1) In tropical Africa, full glacial conditions ended widespread glacier collapse in South America and New early. A "rst warming/wetting phase starting around Zealand (Lowell et al., 1995); a signi"cant increase in 17 ka led the abrupt changes in Greenland by several lake-level in the Bolivian Altiplano prior to the major millenia, but appears coincident with the steepest water rise at ca. 15 ka (Sylvestre et al., 1998); and the temperature increase recognized at Byrd and Vostok in initial warming in the Huascaran ice core in Peru Antarctica. (Thompson et al., 1995). Initial warming in the southern (2) Drastic arid}humid transitions around 15 and hemisphere and tropical regions may have increased the 11.5 ka north of 103S are thought to match the major atmosphere evaporative power, when the North Atlantic Greenland deglacial warming events, in concert with the deep-water (NADW) production and the oceanic ther- switching of the oceanic thermohaline circulation to mohaline circulation were still in glacial mode (Charles modern mode. and Fairbanks, 1992). Thus, water vapour may have been (3) However, part of the signal after 15 ka in southern redistributed over the tropics through the atmospheric and Atlantic equatorial Africa rather seems connected circulation. The 17}16 ka event leads the major deglacial with climate changes at high southern latitudes, as ob- events of the northern high latitudes by about 2000 yr. served in continental Antarctica. This delay could be due to a stationary polar front (4) A synchrony is observed between major wet and located at 453N in the North Atlantic until about 15 ka; dry pulses north of 103S (the largest part of the African it may have insulated areas to the north from the global continent), and rises and falls of the global atmospheric warming (Sowers and Bender, 1995). methane concentration. The amplitude of the hydrologi- A return to dry conditions, poorly dated but between cal steps even follows that of the CH shifts. Bearing in ca. 16 and 15 ka, is identi"able at Madagascar (S6), Lakes mind that most of the modern methane is produced in Victoria (S10) and Albert (S11). This reversal event has tropical wetlands, the step-wise transition toward wetter no clear equivalent in polar ice sheets (Fig. 8). conditions supports the hypothesis of positive feedbacks The warming/wetting events around 15 and 11.5 ka from the African tropics to the global deglacial warming dominate north of 103S, but the former is weak in the through greenhouse gas production (Chappellaz et al., northern tropics and equatorial West Africa. Both 1993; Blunier et al., 1998). roughly match the abrupt temperature rises observed in Greenland ice cores and separated by the YD cold spell, and large and rapid increases in atmospheric methane 5. The Holocene period concentration (Fig. 8). The African events were already associated with the switch to modern mode of NADW Hydrological changes of extremely large amplitude production (Street-Perrott and Perrott, 1990; Gasse occurred during the Holocene in the African tropics and et al., 1990; Gasse and Van Campo, 1994; Broecker et al., subtropics. Although their general envelope is roughly 1998). The reactivation of the oceanic thermohaline con- consistent with orbitally induced variations of the mon- veyor belt at ca. 15 ka, and thus of the Algulhas current, soon strength, changes in e$cient moisture did not sim- may have acted on climates in southeastern Africa and ply respond to the smooth sinewaves of orbital forcing. Madagascar. On the Atlantic side of Africa, the dry interval re- 5.1. Evidence for major Holocene climatic changes corded at 14.4}12.5 ka at site S4, also apparent at S19 and S21, rather appears coincident with the Antarctic In southwest Africa, the Stampriet isotope record (S3) Cold Reversal (ACR; Jouzel et al., 1995; Fig. 8) than with suggests a transition from a dry to a wet climate around climate events in the North Atlantic region. It possibly 7 ka (Fig. 5; Stute and Talma, 1998). The S4 record originated from an increased in#uence of the cold Ben- shows: a very dry interval from 10.9 to 9.3 ka and the guela current along the Atlantic coast. But clearly, the warmest and wettest Holocene period between 6.3 and southern S4-record is in antiphase with the S21-record 4.8 ka (Ning Shi et al., 1998) in phase with increased river after 12.5 ka. discharge and weathering as indicated by clay mineral F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 203

A maximum in aridity is suggested around 4.2 ka at site S6 (Gasse and Van Campo, 1998). North of 103S, the general picture is a high P}E bal- ance during the early-mid Holocene. In equatorial East Africa, maximum P}E balance may have occcurred dur- ing the early Holocene. A diatom record from Lake Victoria (S10; Stager and Mayewski, 1997; Fig. 10) shows a high P/E ratio (or lake level) and good wind-driven mixing from ca. 11 to 7.8 ka, with a maximum between 8.2 and 7.8 ka. An abrupt decrease in wind activity and in lake-level then occurred, in response to enhanced rainfall seasonality and/or a tendency toward aridity. Wind- driven mixing remained low until ca. 3}2 ka; the P/E ratio progressively dropped until today. Hastenrath and Kutzbach (1983) estimated from a water and energy balance model that, at around 9 ka, mean annual rainfall in the basins of lakes Victoria (S10), Naivasha (S13), and Nakuru-Elmenteita (S14) was respectively 20, 15, and 35% higher than today. Lakes Tanganyika (S8), Victoria (S10), Albert (S11) over#owed, and supplied the Congo and the White Nile rivers. Lake Kivu (S9) rose rapidly by 11.5 ka, and over#owed toward Lake Tanganyika at ca. 10.5 ka. This lake remained high up to ca. 4.5 ka, and then dropped. It was closed again from ca 3.5 to 1.3 ka (Haberyan and Hecky, 1987). A lowering of water level took place at Lake Turkana (S15) from ca. 5 ka (Fig. 7) and a closed-basin lake status was achieved permanently at ca. 4.2 ka (Johnson, 1996). In equatorial West Africa, maximum wetness occcur- red during the mid-Holocene, from about 8}7.5 to 4 ka; drier conditions then established and culminated around 2.2 ka, before a return to wetter conditions around 1.5}1 ka (Maley, 1991,1997; Maley and Brenac, 1998). Fig. 10. Changes in limnological conditions at Lake Victoria over the Lake Bosumtwi (S21) experienced marked drops in lake past 11 ka as inferred from diatom assemblages (b and c), compared level centred around 8.3}8 ka, 4.2}4, and 0.5 ka (Talbot with windblown dust concentration in the GISP ice core in Greenland et al., 1984; Fig. 9). (a), and in the Taylor Dome ice core in Antarctica (d). From Stager and In the Sahelian belt, the Bahr-el-Ghazal depression Mayewski (1997). Dust concentration after O'Brien et al. (1995), and (S23), which received the out#ow of Lake Chad, and the Mayewski et al. (1995). Ethiopian lakes (S16}S17) experienced an evolution very similar to that of Lake Bosumtwi (Fig. 9), especially in the data (Gingele, 1996). In South Africa, dry/cool conditions timing of short-term dry events around 8.3}8 ka and prevailed during the early Holocene (Scott, 1989,1993; 4.2}4 ka. In response to dry conditions in both Ethiopia Tyson, 1991; Partridge, 1997), and `optimala Holocene and equatorial Africa, the Main Nile #ood was consider- temperatures from latitudes 23}283S occurred at 8}7ka ably reduced around 4.2 ka (Said, 1993). A minor dry (Scott, 1993). At Lake Malawi (S7), an early Holocene spell is also recorded at 7}6.5 ka in Ethiopia (Fig. 9) and low lake-level at !100 m, !150 m (Finney et al., 1996; in several Sahelian sites (e.g., S26}S27; Gasse and Van Fig. 7) may correspond to a decrease in precipitation of Campo, 1994). High resolution records of Bal Lake and ca. 29}25%, according to simulations based on the mod- Kajemarum Oasis (S25; Holmes et al., 1999) show that ern hydrological budget of this lake (Bergonzini, 1997). the early mid-Holocene organic silt accumulation, which During the late Holocene, cooler/drier conditions in re#ects wet conditions, was interrupted by eolian sand the northern Kalahari occurred between 4.8 and 3.3 ka layers at 8.1 and 4.1 ka; a marked climate deterioration (Ning Shi et al., 1998). A temperature decrease in the started at ca. 4.1 ka with an extremely severe drought at Mediterranean belt of South Africa was also inferred 1.2}1.0 ka. between 4.5 and 2 ka from stable isotope data from An early-mid-Holocene `greena Sahara is evidenced Elands bay (Cohen et al., 1992), and around 4.5 ka from by numerous palaeolake, palaeodrainage, vegetation, speleothems of Cango Cave (Talma and Vogel, 1992). and archeological records (see, e.g., Hillaire-Marcel et al., 204 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

Early-mid-Holocene waters of northern Mali and north- ern Niger (e.g. Fontes et al., 1993; Dodo and Zuppi, 1997), Sudan (Darling et al., 1987) and Libya (Edmunds and Wright, 1979) di!er from late Holocene and modern rainfalls, especially in their very low deuterium excess (Fig. 11) for which no satisfactory explanation has yet been proposed. Their very low dO-values (dO: !10& to !12&) were attributed to convective showers gener- ated by squall lines that develop along the ITCZ (Fontes et al., 1993). This hypothesis implies an ITCZ northwards movement of at least 500 km, in agreement with palaeolake data. Such a migration of Sahelian environ- ments is con"rmed by the high dissolved nitrogen con- tent in Holocene groundwater (up to 45 mg kg\ of nitrate in northern Mali), re#ecting biologically active Fig. 11. Groundwater stable isotope composition from the semi-con- "ned aquifer complex of the `Route du Sela in northern Mali, as an soils (Fontes et al., 1993). example of Holocene paleowaters in the Sahara. Comparison with modern water from the Niger River, and recently in"ltrated waters 5.2. The greening of the Sahara close to Tombouctou. GMWL: Global Meteoric Water Line. After Fontes et al. (1993). Climate and vegetation simulations have been used to examine the in#uence on the African monsoon, of Earth orbital changes, SST changes, and feedbacks associated 1983, Baumhauer, 1991; Fontes and Gasse, 1991; Petit- with changes in soil moisture at 6 ka, but most models Maire et al., 1993; Pachur and Altmann, 1997; Hoel- fail to simulate conditions much wetter than today in the zmann et al., 1999; Cremaschi and Di Lernia, 1998, for western Sahara: in simulations by AGCMs (Kutzbach regional syntheses). Between 17 and 183N, a savanna et al., 1993; de Noblet et al., 1996; Jolly et al., 1998), ecosystem was maintained without interruption during a coupled ocean-atmosphere model (Kutzbach and Liu, the early and mid-Holocene, when parts of the large, now 1997), and a coupled atmosphere-biome model (Texier arid, endorheic region of Western Nubia drained into the et al., 1997), the amplitude of the climate change appears Nile via the Lower Wadi Howar (Fig. 3; Kroepelin, 1993). underestimated. Recently, a synchronously coupled at- In western Fezzan (S33), a wet savannah landscape inside mosphere}ocean}vegetation model showed that changes the mountains and shallow lakes in the ergs developed in vegetation cover during the mid-Holocene modify and from ca. 10 to 6 ka, connecting the Sahelian area to the amplify the climate system's response to an enhanced northern Sahara; a dry period at 8.5}8 ka interrupted this seasonal cycle of insolation in the northern hemisphere wet interval (Cremaschi and di Lernia, 1998). In the (Ganopolski et al., 1998b). This model results show western Sahara, the northern fringe of the palaeo-Sahel strong synergetic e!ect of changes in vegetation cover, has reached about 223N (Petit-Maire and Riser, 1993). ocean temperature, and sea-ice at boreal latitudes, but Some depressions (e.g., S28) experienced two lacustrine the atmosphere}vegetation feedback is most important episodes between 10 and 4.5 ka separated by a long dry in the subtropics. The model simulates increased precipi- phase centred around 8 ka; others remained dry after tation over the whole northern Africa, and thus provides 8 ka (e.g., S29). Permanent lakes also occurred along the a better level of agreement with geological data from northern margin of the Sahara during the early-mid- northwest Africa. It should be noted, however, that Holocene: the Sebkha Mellala (S35) experienced two freshwater lacustrine episodes between 10.3 and 5.7 ka, (1) The 6 ($0.5) ka snapshot which was selected for separated by a desiccation phase at ca. 8.5-8.0 ka (Gasse simulations makes the model-data comparison di$cult et al., 1990); shallow waterbodies have occurred at site in Africa, where this time slice falls during a period of S36 from ca. 11}10 ka to 3.3 ka (Gasse et al., 1987), while climate instability. high lake-level persisted until today in the Atlas Moun- (2) Many of the early-mid-Holocene waterbodies of tains of Morocco (Lamb et al., 1995). Dry conditions the Sahara-Sahel were the surface expression of water- prevailed after 6}5.5 ka in the eastern Sahara, and table rises in the aquifers. Waterbodies and soil moisture 4.5}4 ka in the whole Sahara, although a further late in the lowlands were generated by local precipitation, but Holocene wet episode of minor amplitude is recorded at may have also resulted from groundwater #ows from several places (Maley, 1997). General aridity took place remote highlands. Furthermore, a delay may have occur- at ca. 2 ka. red between the onset of arid conditions and the "nal The wet Holocene period has also been a period of desiccation of lowlands. For instance, in the Nubian intensive recharge of the Saharan deep water reservoirs. aquifer system, the very low rate of drawdown "ts the F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 205 idea of continuously decreasing groundwater levels since throughout much of the eastern two-thirds of the United the recharge period, about 7 ka. The hydraulic head States (Bradbury and Dean, 1993). In the African north- decreased to a regional mean of 60 m below the ground ern tropics and in the northern Sahara, a dry spell cen- surface, but groundwater is still fairly close to ground tred at 8.5}8 ka interrupted a generally wet climate surface in the depression areas (Christman and Sonntag, episode. This short-life dry event, apparently in phase 1987). with a weak monsoon episode in the Indian monsoon (3) Today, the Sahara north of 20}243N belongs to the domain of West Asia (Gasse and Van Campo, 1994), is `winter rainfall patterna, while precipitation in the south- followed by a return to `initiala conditions, as the cool ern Sahara is associated with the summer monsoon spell that a!ected high latitudes of the northern hemi- (Fig. 2). Hydrological changes at sites at 30}323N pres- sphere. ently supplied by winter precipitation on the Atlas and A dry event in the northern tropics around 7}6.5 ka the northern Sahara heights can hardly be attributed to was recorded in Africa (e.g., S16, S16, S26, S27) and in a monsoonal climate. It is not possible yet to determine West Asia (Gasse and Van Campo, 1994). In western the seasonal distribution of past precipitation, and thus Tibet, dO-records of lacustrine carbonates suggest that to identify the sources of water vapour responsible for this event marks the start of the monsoon retreat and `greeninga in the northern Sahara. a change in the origin of air moisture, from monsoonal to locally re-evaporated moisture under drier conditions 5.3. Holocene hydrological and climatic instability (Wei and Gasse, 1999). As with the ca. 8.2 ka event, climate changes at the mid Attention is paid here to some major events apparent to late Holocene transition, ca. 4.5}4 ka, either corre- at numerous sites around 8.5}7.8, 7}6.6 and 4.5}3.5 ka, sponded to a short-term dry/cool episode, or inaugur- and which have a!ected regions well beyond the tropics ated new atmospheric, climatic conditions. At many sites and subtropics of Africa. These events have di!erent (S9, S15}17, S21, S23, S27, S28), the 4.5}4 ka event is regional expressions. They are either a step-over shift to a dry spell followed by a further wet phase of low ampli- a new millenial-duration background climate, or tude. It coincides with an unusual dust event identi"ed at a short-lived #uctuation rapidly followed by a return 4.2 ka in the eastern Mediterranean and across Meso- to initial conditions. potamia, from Lake Van to the Gulf of Oman (Dalfes et Rapid climate changes occurring in the interval al., 1997), and with the episode of maximum Holocene 8.5}7.8 ka probably correlate with the most prominent aridity in Western Tibet (Gasse and Van Campo, 1994). Holocene climatic event observed in Greenland ice-cores, A trend toward aridity, culminating around 2.2 ka, and recorded there by the only distinct Holocene started at ca. 4.5}4 ka in equatorial West Africa. O shift and a signi"cant decrease in methane at ca. Cooler/drier climates took place in Southwest Africa 8.4}8.0 ka (Chappellaz et al., 1993; Alley et al., 1997). The (S4). The changes were opposite to those of the early- Greenland event "ts with one of the Holocene cool, mid-Holocene transition, the 4.5}4 ka event coinciding ice-bearing water episodes (nr 5) in the North Atlantic with the re-establishment of drier and warmer conditions (Bond et al., 1997). It is coincident with a peak in aerosol in the West Mediterranean basin (Kallel et al., 1998), the concentrations in polar ice cores from Greenland and onset of generally cooler climates and increased moisture Antarctica. This peak was followed by an abrupt shift at in the north-central United States (Bradbury and Dean, 7.8 ka (Fig. 10). After 7.8 ka, windblown chemical indi- 1993), increased wetness in Amazonia and Central Andes cators recovered quickly in Greenland ice, while they leading to drastic lake-level rises of Lake Titicaca (Mar- remained low in Antarctica. tin et al., 1993; Mourguiart et al., 1998). In South Amer- At Lake Victoria (Fig. 10), the 8.5}7.8 ka event led to ica, this transition toward wetter conditions was a long period of decreasing P}E balance that maintained interpreted as a change from warm environments with- until today. Stager and Mayewski (1997) concluded that out El Nin o episodes to the present cooler climate inter- Holocene climatic #uctuations in equatorial East Africa rupted by El Nin o every few years (Markgraf et al., 1991). are linked to changes in Antarctica weather and circum- The causes of these abrupt Holocene events are not polar ocean currents. In subequatorial Atlantic Africa, known. Alley et al. (1997) hypothesized a decrease in e.g. at Lake Bosumtwi (Fig. 9), a 8.5}7.8 ka dry event North Atlantic thermohaline circulation around 8.2 ka preceeded the episode of maximum P!E that lasted caused by an increase in freshwater #ux to the North over about 4000 yr. It thus corresponds with the onset of Atlantic. Gasse and Van Campo (1994) noted that the speci"c mid-Holocene climate conditions, as at many weak monsoon episodes around 8 and 4 ka are roughly places in the world: generally cooler, wetter conditions in coincident with decreases in North Atlantic SST and the western Mediterranean basin (Kallel et al., 1998); very surface salinity. The antiphase between rainfall changes low e!ective moisture in the Amazon basin and in the in Equatorial West Africa and in the Amazon may re#ect Chilian and Bolivian Altiplano (Valero-Garce`s et al., changes either in the Walker circulation and ENSO 1996); reduced precipitation and increased windiness phenomena, or in the Hadley circulation, e.g. a strength- 206 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 ening of the northern trade winds at ca. 4 ka, bringing cycle will be required from both hemispheres to estimate dry, harmattan wind #ows in Africa, and water vapour the relative roles of the Milankovitch cycles and of from the Atlantic in South America. Whatever their ocean}land interactions on African climates under gla- cause, the worldwide impacts of these events imply that cial and interglacial conditions. Potential sites to obtain rapid changes in the water cycle in Africa are the expres- long continuous records are the large Rift lakes of East sion of global climatic reorganizations. Africa and numerous crater lakes from East and equato- rial West Africa, and Madagascar. Archives for remote periods are rare in arid regions because of the wind 6. Conclusions de#ation, but groundwaters and speleothems are promising. The palaeohydrological evidence collected here yields In fact, over the period for which radiocarbon dating is some insights on the sensitivity of the African climate to possible, the hydrological history of Africa appears to changes in global climate forcing factors. It also raises have been a complex alternation of wet and dry episodes a series of questions that remain to be solved. The main with abrupt transitions, which cannot be directly ac- points arising from the palaeohydrological evidence are counted for by orbital forcing. As already highlighted for as follows. the last deglaciation (Broecker et al., 1998), the Earth's climate jumped from one mode of operation to another (1) As a whole, climate information in Africa su!ers rather than proceeding in smooth sinewaves. These large geographical and methodological gaps. First, the abrupt changes re#ect reorganizations in the transport of climate evolution in the southern tropics is poorly energy and water vapour at the global scale. The few documented and the large Congo basin is practically records covering the past 40,000 yr (e.g., S6-record) show unknown. Second, whatever the region, most records are that the glacial period has been punctuated by short-term undermined by low time resolution and large dating warm/dry pulses. Further work is needed to understand uncertainties. Third, African palaeohydrological data are if theses pulses were related to the interstadial warm mainly qualitative while model-data comparisons require events in the GRIP ice cores and/or with SST variations quantitative estimates of past hydroclimatic variables. in the surrounding tropical oceans. Further studies are needed to estimate palaeo-temper- (3) During the LGM, most records indicate generally ature, palaeo-precipitation and palaeo-evaporation from dry conditions in both hemispheres, while orbital forcing noble gas groundwater records, speleothem isotope re- predicts enhanced summer rainfall in the southern cords, and modelling of individual lake systems combin- tropics. Geological evidence presented here roughly ing water, salt, isotope, and energy balance equations. agrees with recent coupled ocean}atmosphere climate Fourth, multi-proxy and multi-archive studies need to be models which simulate cooler conditions and decreased developed further. Recent works (e.g. Cremaschi and Di summer rainfall in the tropics. Lower tropical sea-surface Lernia, 1998; Holmes et al., 1999) have shown that the conditions have played a role on tropical climates by combination of surface water (lakes, rivers), groundwater reducing the #ux of water vapour inland. (aquifers, speleothems) and sand dune studies leads to We need, however, to obtain a better picture of re- better understand the system evolution. Such approaches gional LGM climates to fully understand the impact of should be made whenever possible, especially when changes in atmospheric circulation and oceanic currents di!erent dating techniques can be applied, allowing along the eastern and western coasts of Africa. Because of a cross-control of the chronological framework. Fifth, the possible complexity of the 23}18 ka interval, model- correlations between inland and marine records, poorly data comparison should be based on a chronology more developed so far, will be required to understand accurate than that available in many records. More im- ocean}continent interactions. Marine cores close to the proved paleoclimatic data, and more simulations with major African rivers mouth may help establish such detailed models will be required to arrive at a robust correlations. quantitative understanding of hydrological changes in (2) At long-time scale, orbitally induced changes in the tropics associated with a global hydrological cycle summer solar radiation certainly account for a large part weaker than today. of hydrological changes in the tropics. The orbital pre- (4) The "rst signs of the deglaciation in the African cession appears to be a satisfactory explanation for the tropics and subtropics started around 17 ka, much earlier cyclicity in summer precipitation changes in the Pretoria than the major deglacial events observed in the North Saltpan record (South Africa) from 200 to 50 ka, but the Atlantic regions. Initial warming in the southern hemi- signal is blurred during the LGM. Very few long records sphere may have a!ected the water cycle in the entire are available. No comparison is possible between the African continent at ca. 17}16 ka through changes in the penultimate and the last glaciations, or between the atmospheric circulation. North of 103S, sudden and very Holocene and the last interglacial periods. Other large increases in the water budget at ca. 15 and 11.5 ka palaeoclimatic sequences spanning at least one climatic were coincident with the onset of the BoK lling-Allerod and F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 207

Pre-Boreal warm intervals identi"ed in northern high range of natural climatic variability of the current inter- latitudes. It was the reactivation of the oceanic ther- glacial period. There is no guarantee that similar shifts mohaline circulation that was apparently instrumental in with serious social impact cannot happen again. We need bringing the interglacial shift to the continents. to estimate their amplitude, to better constrain their Linkages between high southern latitudes and tropical timings, to obtain high resolution Holocene records from Africa climates apparently maintained after 15 ka. The the surrounding oceans, before concluding on their rela- dry interval from 14.5}14 to 12.5 ka in southern tions with short-term events known elsewhere (e.g. in the and equatorial West Africa may be related to the ACR. North Atlantic region) and to analyze potential mecha- Holocene climatic #uctuations in equatorial East Africa, nisms. An understanding of natural #uctuations in water e.g. at Lake Victoria, also appear to have been linked to resources that occur at the time scale of human lives the Antarctica climate. More detailed marine and conti- should be a "rst priority in future research. nental records in the tropics are needed to reconcile low (7) Palaeohydrological data summarized in this paper and high latitude climate records of both hemispheres. show that Africa has experienced rapid changes in the (5) Once the glacial boundary conditions decayed, continental water cycle of considerable magnitude, re- generally wet climate established in the northern tropics gardless of the presence or absence of large polar ice while the southern tropics became drier during the sheets. Because water vapour is the most powerful green- early Holocene, in response to orbital forcing. However, house gas, large hydrological changes in the tropics may changes in summer insolation alone cannot account for have signi"cant feedback e!ects on global climate. Fur- some regional climate patterns, e.g., conditions much ther paleohydrological research should aim at obtaining wetter than today in the northwestern Sahara around solid chronologies, but also at analysing the mechanisms 6 ka, when the western Mediterranean basin was colder of water storage and losses in surface and groundwater and wetter than today and when the eastern Sahara was reservoirs, obtaining quantitative reconstructions of hy- becoming drier. Climate models showed that such a wet- drological cycles, and identifying moisture transport pat- ting induced complex interactions and feedback pro- terns at regional scales. Together with the development cesses between the atmosphere, the oceans, and of climate-system models with "ner spatial and temporal vegetation cover. However, the origin of water vapour scales, this will allow a better understanding of how over the Sahara is not identi"ed. Stable isotope studies of hydrological changes in the tropics respond to changes Holocene groundwater along North}South and in major-forcing factors, and how the tropics in#uence East}West transects could bring important insights on climate globally. the moisture transport patterns. In this region where many waterbodies were supplied by aquifers, the water cycle could be better understood by combining quanti- References "ed surface and groundwater data, and by modelling the entire hydrological systems. Annually laminated sedi- Adams, L.J., Tetzla!, G., 1985. The extension of Lake Chad at about ments from lakes or speleothems may also provide in- 18,000 yr BP. Zeitschrift fuK r Gletscherlunke und glazialgeologie 21, formation about the rainfall season. 115}123. (6) Until recently, the Holocene climate was com- Alley, R.B., Mayewski, P.A., Sowers, T., Stuiver, M., Taylor, K.C., monly thought to have been fairly stable. This idea was Clark, P.U., 1997. Holocene climatic instability: a prominent, wide- spread event 8200 yr ago. Geology 25 (6), 483}486. mainly based on polar ice and North Atlantic deep-sea Ariztegui, D., Bianchi, M.M., Masaferro, J., Lafargue, E., Niessen, F., core records where Holocene #uctuations are indeed of 1997. Interhemispheric synchrony of Late-glacial climatic instabil- very low amplitude compared to those of the glacial- ity as recorded in proglacial Lake Mascardi. Argentina. Journal of deglacial transition. On the contrary, dramatic hy- Quaternary Science 12 (4), 333}338. drological changes have long been apparent in Africa, Bard, E., Rostek, F., Sonzogni, C., 1997. Interhemispheric synchrony of the last deglaciation inferred from alkenone palaeothermometry. and appear as large in magnitude as their glacial counter- Nature 385, 707}710. parts. Dramatic changes in water resources have enor- Baumhauer, R., 1991. Palaeolakes of the south central Sahara: prob- mous consequences on human populations, generating lems of palaeoclimatological interpretation. Hydrobiologia 214, famines, migrations, civilization foundations and collap- 347}357. ses. The advanced urban civilizations in Egypt Hassan, Beck, J.W., ReH cy, J., Taylor, F., Edwards, R.L., Cabioch, G., 1997. Abrupt changes in early Holocene tropical sea surface temperature 1997, Mesopotamia and India mysteriously collapsed derived from coral records. Nature 385, 705}707. around 4000 yr ago (Dalfes et al., 1997). This collapse of Berger, A.L., 1978. Long-term variations of daily insolation and Quat- the Old Word societies is likely to be related to the ernary climatic changes. Journal of Atmospheric Science 35, 4.2}4 ka dry event that a!ected the monsoon domain and 2362}2367. eastern Mediterranean. There is no doubt that Holocene Bergonzini, L., 1997. Bilans hydriques de lacs du Rift Est-Africain (Kivu, Tanganyika, Rukwa et Nyassa). Approche mensuelle et an- drought and #ood episodes have been more dramatic nuelle. Essai d'interpreH tation de la variabiliteH inter-annuelle et des and more persistent than those of the instrumental re- #uctuations passeH es. Thesis, UniversiteH Paris-Sud, Orsay, France, cord of the last century, which does not represent the full pp. 246. 208 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

Bergonzini, L., ChalieH , F., Gasse, F., 1997. Paleoevaporation and Cockcroft, M.J., Wilkinson, M.J., Tyson, P.D., 1987. The application of paleoprecipitation in the Tanganyika Basin at 18,000 yr B.P. infer- a present-day climatic model to the Late Quaternary in southern red from hydrologic and vegetation proxies. Quaternary Research Africa. Climate Change 10, 161}181. 47, 295}305. Cohen, A.L., Parkington, J.E., Brundrit, G.B., Van de Merwe, N.J., Beuning, K.R.M., Talbot, M.R., Kelts, K., 1997. A revised 30,000-year 1992. A Holocene marine climate record in mollusc shells from the paleoclimatic and paleohydrologic history of Lake Albert, East southwest African coast. Quaternary Research 38, 379}385. Africa. Palaeogeography, Palaeoclimatology, Palaeoecology 136, Dalfes, H.N., Kukla, G., Weiss, H. (Eds.), 1997. Third Millenium BC 259}279. climate change and old world collapse. NATO ASI Series. Series 1, Blunier, T., Chappellaz, J., Schwander, J., DaK llenbach, A., Stau!er, B., Vol. 49. pp. 723. Stocker, T.F., Raynaud, D., Jouzel, J., Clausen, H.B., Hammer, C.U., Darling, W.G., Edmunds, W.M., Kinniburgh, D.G., Kotoub, S. et al., Johnsen, S.J., 1998. Asynchrony of Antarctic and Greenland climate 1987. Sources of recharge to the basal Nubian Sandstone aquifer, change during the last glacial period. Nature 394, 739}743. Butana region, Sudan. In: Isotope Techniques in Water Resources Bond, G.C., Heinrich, H., Broecker, W., Labeyrie, L., McManus, J., Development. IAEA, Vienna, Proceedings Series STI/PUB/757, pp. Andrews, J., Huon, S., Jantschik, B., Clasen, S., Simet, C., Tedesco, 205}224. K., Klas, M., Bonani, G., Ivy, S., 1992. Evidence for massive dis- deMenocal, P.B., Rind, D., 1996. Sensitivity of subtropical African and charges of icebergs into the North Atlantic ocean during the last Asian climate to prescribed boundary conditions changes: model glacial period. Nature 360, 245}249. implications for the Plio-Pleistocene evolution of low-latitude cli- Bond, G.C., Showers, W., Cheseby, M., Lotti, R., Almasi, P., de- mate. In: Johnson, T.C., Odada, E.O. (Eds.), The Limnology, Clima- Menocal, P., Priore, P., Cullen, H., Hadjas, I., Bonani, G., 1997. tology and Paleoclimatology of the East African Lakes. Gordon A pervasive millenial-scale cycle in North Atlantic Holocene and and Breach, Amsterdam, pp. 57}77. glacial climate. Science 278, 1257}1266. de Noblet, N., Braconnot, P., Joussaume, S., Masson, V., 1996. Sensitiv- Bonne"lle, R., ChalieH , F., Guiot, J., Vincens, A., 1992. Quantitative ity of simulated Asian and African summer monsoons to orbitally estimates of full glacial temperatures in equatorial Africa from induced variations in insolation 126, 115 and 6 k BP. Climate palynological data. Climate Dynamics 6, 251}257. Dynamics 12, 589}603. Bradbury, J.P., Dean, W.E. (Eds.), 1993. Elk Lake, Minnesota: evidence Dodo, A., Zuppi, G.M., 1997. Etude des eH coulements souterrains dans for rapid climate change in the North-Central United States. U.S. le bassin de Bilma-Djado a` l'aide des isotopes de l'environnement. Geological Survey, Special Paper 276, 309}328. Comptes Rendus del' AcadeH mie des Sciences, Paris, SeH rie D 325, Broecker, W.S., Peteet, D., Hajdas, I., Lin, J., Clark, E., 1998. Antiphas- 845}852. ing between rainfall in Africa's Rift Valley and North America's Durand, A., 1993. Enregistrement seH dimentaire de la dynamique Great Basin. Quaternary Research 50, 12}20. climatique au Quaternaire supeH rieur dans le Sahel Central (Niger Bush, A.G.B., Philander, S.G.H., 1998. The role of ocean-atmosphere et Tchad). Thesis, UniversiteH de Bourgogne, Dijon, France, interactions in tropical cooling during the Last Glacial Maximum. pp. 607. Science 279, 1341}1344. Edmunds, W.M., Fellman, E., Baba Goni, I., 1999. Environmental Carrara, C., Cremaschi, M., Quinif, Y., 1998. The travertine deposits in change, lakes and groundwater in the Sahel of Northern Nigeria. the Tadrart Acacus (Libyan Sahara nature and age. In: Cremaschi, Journal of the Geological Society, London 156, 345}355. M., Di Lernia, S. (Eds.), Wadi Teshuinat palaeoenvironment and Edmunds, W.M., Wright, E.P., 1979. Groundwater recharge and prehistory in south-western Fezzan (Libyan Sahara). All'Insegna del paleoclimate in the Sirte and Kufra basins, Libya. Journal of Hy- Giglio, CNR, Milano, pp. 59}66. drology 40, 215. Causse, C., Conrad, G., Fontes, J.-C., Gasse, F., Gibert, E., Kassir, A., Finney, B.P., Sholtz, C.A., Johnson, T.C., Trumbore, S., 1996. Late 1988. Le dernier `Humidea du PleH istoce`ne du Sahara nord-occiden- Quaternary lake-level changes of Lake Malawi. In: Johnson, T.J., tal daterait de 80}100 000 ans. Comptes Rendus del' AcadeH mie des Odada, E.O. (Eds.), The Limnology, Climatology and Paleoc- Sciences, Paris, SeH rie 2 306, 1459}1464. limatology of the East African lakes. Gordon and Breach, London, ChalieH , F., 1995. PaleH oclimats du bassin Tanganyika Sud au cours des pp. 495}508. 25 derniers mille ans. Reconstruction quantitative par le traitement Fontes, J.Ch., Gasse, F., 1991. PALHYDAF (Paleohydrology in Africa) statistique des donneH es polliniques. Comptes Rendus del' AcadeH mie Program: objectives, methods, results. Palaeogeography, Palaeoc- des Sciences, Paris, SeH rie 2 320 (2), 205}208. limatology, Palaeoecology 84, 191}215. Chappellaz, J., Blunier, T., Raynaud, D., Barnola, J.M., Schwander, J., Fontes, J.Ch., Gasse, F., Andrews, J.N., 1993. Climatic conditions Stau!er, B., 1993. Synchronous changes in atmospheric CH of Holocene groundwater recharge in the Sahel zone of Africa. and Greenland climate between 40 and 8 kyr BP. Nature 366, In: Isotope Techniques in the Study of Past and Current Environ- 443}445. mental Changes in the Hydrosphere and the Atmosphere. IAEA- Charles, C., Fairbanks, R., 1992. Evidence from Southern Ocean sedi- SM-329/59, International Atomic Energy Agency, Vienna. ments for the e!ect of North Atlantic deep-water #ux on climate. Ganopolski, A., Rahmstorf, S., Petoukhov, V., Claussen, M., 1998a. Nature 355, 416}419. Simulation of modern and glacial climates with a coupled global Clemens, S.C., Prell, W.L., 1991. Late Quaternary forcing of Indian model of intermediate complexity. Nature 391, 351}356. Ocean summer-monsoon winds: a comparison of Fourier model Ganopolski, A., Kubatzki, C., Claussen, M., Brovkin, V., Petouhkov, and circulation model results. Journal of Geophysical Research 96 V., 1998b. The in#uence of vegetation-atmosphere-ocean interac- (12), 22683}22700. tion on climate during the Mid-Holocene. Science 280, 1916}1919. CLIMAP Members, 1981. Seasonal Reconstruction of the Earth's Sur- Gasse, F., 1977. Evolution of Lake AbheH (Ethiopia and T.F.A.I.) from face at the Last Glacial Maximum. Geological Society of America 70,000 B.P. Nature 2, 42}45. Map Chart Ser. MC-36. Gasse, F., Fontes, J.Ch., 1992. Climatic changes in northwest Africa Christmann, D., Sonntag, C., 1987. Groundwater evaporation from during the last deglaciation. In: Bard, E., Wallace, W.S. (Eds.), The East-Saharian depressions by means of deuterium and oxygen-18 in Last Deglaciation: Absolute and Radiocarbon Chronologies. Nato soil moisture. Isotope Techniques in Water Resources Develop- ASI Series, Vol. 12. Springer, Berlin, pp. 295}325. ment. IAEA, Vienna, IAEA-SM-299/37, pp. 189}204. Gasse, F., Fontes, J.Ch., Plaziat, J.C., Carbonel, P., Kaczmarska, I., De Cremaschi, M., Di Lernia, S. (Eds.), 1998. Wadi Teshuinat Palaeoen- Deckker, P., SoulieH -Marsche, I., Callot, Y., Dupeuple, P.A., 1987. vironment and Prehistory in South-Western Fezzan (Libyan Sa- Biological remains, geochemistry and stable isotopes for the hara). All'Insegna del Giglio, CNR, Milano, pp. 332. reconstruction of environmental and hydrological changes in the F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 209

Holocene lakes from North Sahara. Palaeogeography, Palaeo- J.P., 1992. Irregular glacial interstadials recorded in a new Green- climatology, Palaeoecology 60, 1}46. land ice core. Nature 359, 311}313. Gasse, F., LeH deH e, V., Massault, M., Fontes, J.Ch., 1989. Water-level Johnsen, S.J., Dansgaard, W., Clausen, H.B., Langway Jr., C.C., 1972. #uctuations of Lake Tanganyika in phase with oceanic changes Oxygen isotope pro"les through Antarctic and Greenland ice during the last glaciation and deglaciation. Nature 342, 57}69. sheets. Nature 235, 429}434. Gasse, F., Street, F.A., 1978. Late Quaternary lake-level #uctuations Johnson, T.C., 1996. Sedimentary processes and signals of past climate and environments of the northern rift valley and Afar region change in the large lakes of East African Rift Valley. In: Johnson, (Ethiopia and Djibouti). Palaeogeography, Palaeoclimatology, T.C., Odada, E.O. (Eds.), The Limnology, Climatology and Paleoc- Palaeoecology 25, 145}150. limatology of the East African Lakes. Gordon and Breach, Amster- Gasse, F., TeH het, R., Durand, A., Gibert, E., Fontes, J.Ch., 1990. The dam, pp. 367}412. arid-humid transition in the Sahara and the Sahel during the last Johnson, T.C., Scholz, C.A., Talbot, M.R., Kelts, K., Ricketts, R.D., deglaciation. Nature 346, 141}156. Ngobi, G., Beuning, K., Ssemmanda, I., McGill, J.W., 1996. Late Gasse, F., Van Campo, E., 1994. Abrupt post-glacial climate events in Pleistocene desiccation of Lake Victoria and rapid evolution of West Asia and North Africa monsoon domains. Earth and Planet- Cichlid "shes. Science 273, 1091}1093. ary Science Letters 126, 435}456. Jolly, D., Haxeltine, 1997. E!ect of low atmospheric CO on tropical Gasse, F., Van Campo, E., 1998. A 40,000 yrs pollen and diatom record african montane vegetation. Science 276, 786}788. from the southern tropics (Lake Tritrivakely, Madagascar Pla- Jolly, D., Harrison, S.P., Damnati, B., Bonne"lle, R., 1998. Simulated teaux). Quaternary Research 49, 299}311. climate and biomes of Africa during the Late Quaternary: compari- Gillespie, R., Street-Perrott, F.A., Switzur, R., 1983. Post-glacial arid son with pollen and lake status data. Quaternary Science Reviews episodes in Ethiopia have implications for climate prediction. Na- 17, 629}657. ture 306, 680}683. Jouzel, J., Lorius, C., Petit, J.R., Genthon, C., Barkov, N.I., Kotlyakov, Gingele, F.X., 1996. Holocene climatic optimum in southwest Africa- V.M., Petrov, V.M., 1987. Vostok ice core: a continuous isotope evidence from the marine clay mineral record. Palaeogeography, temperature record over the last climatic cycle (160,000 years). Palaeoclimatology, Palaeoecology 122, 77}87. Nature 329, 403}408. Guendouz, A., Moulla, A., Edmunds, W.M., Shand, P., Poole, J., Zouari, Jouzel, J., Vaikmae, R., Petit, J.R., Martin, M., Duclos, Y., Stievenard, K.,Mamou,A.,1998.Paleoclimaticinformationcontainedingroun- M., Lorius, C., Toots, M., MeH lie`res, M.A., Burckle, I.H., Barkov, dwaters of the Grand Erg Oriental, North Africa. In: Isotope Tech- N.I., Kotlyakov, V.M., 1995. The two-step shape and timing of the niques in the Study of Past and Current Environmental Changes in last deglaciation in Antarctica. Climate Dynamics 11, 151}161. the Hydrosphere and the Atmosphere. Proceedings of Vienna Sym- Kallel, N., Paterne, M., Duplessy, J.C., Vergnaud-Grazzini, C., Pujol, posium 1997, IAEA, Vienna, SM-349/43, pp. 555}571. C., Labeyrie, L.D., Arnold, M., Fontugne, M., Pierre, C., 1998. Guilderson, T.P., Fairbanks, R.G., Rubenstone, J.L., 1994. Tropical Enhanced rainfall on Mediterranean region during the last sapropel temperature variations since 20,000 years ago: modulating inter- event. Oceanologica Acta 20, 697}712. hemispheric climate change. Science 263, 663}665. Kroepelin, S., 1993. Late Quaternary environments at the Lower Wadi Haberyan, K.A., Hecky, R.E., 1987. The late Pleistocene and Holocene Howar (Southeastern Sahara/NW Sudan). Berliner Geographische stratigraphy and paleolimnology of Lakes Kivu and Tanganyika. Abhandlungen 54, 1}193. Palaeogeography, Palaeoclimatlogy, Palaeoecology 61, 169}197. Kutzbach, J.E., Guetter, P.J., Behling, P.J., Delin, R., 1993. Simulated Hassan, F.A., 1997. Nile #oods and political disorder in Early Egypt. In: climatic changes: results of the COHMAP climate-model experi- Third Millenium BC Climate Change and Old World Collapse, ments. In: Global Climates Since the Late Glacial Maximum. Uni- NATo ASI Series. Series 1, Vol. 49, pp. 1}24. versity of Minnesota Press, Minneapolis, pp. 5}11. Hastenrath, S.L., Kutzbach, J.E., 1983. Paleoclimate and water budget Kutzbach, J.E., Liu, Z., 1997. Response of the African monsoon to of East African lakes. Quaternary Research 19, 141}153. orbital forcing and ocean feedbacks in the Middle Holocene. Hillaire-Marcel, C., Riser, C., Rognon, P., Petit-Maire, N., Rosso, J.C., Science 278, 440}443. Soulie-Marsche, I., 1983. Radiocarbon chronology of Holocene Kutzbach, J.E., Street-Perrott, F.A., 1985. Milankovitch forcing of hydrologic changes in Northeastern Mali. Quaternary Research 20, #uctuations in the level of tropical lakes from 18 to 0 kyr BP. Nature 145}164. 317, 130}134. Hoelzmann, P., Jolly, D., Harrison, S.P., Laarif, F., Bonne"lle, Lamb, H.F., Gasse, F., Ben Kaddour, A., El Hamouti, N., Van der R., Pachur, H.J., 1998. Mid Holocene land-surface conditions in Kaars, S., Perkins, W.T., Pearce, N.J., Roberts, C.N., 1995. Relation northern Africa and the Arabian peninsula: a data set for the between century-scale Holocene arid intervals in tropical and tem- analysis of biogeophysical feedbacks in the climate system. Global perate zones. Nature 373, 134}137. Biogeochemical Cycles 12, 35}41. Lowell, T.V., Heusser, C.J., Andersen, B.G., Moreno, P.I., Hauser, A., Holmes, J.A., Street-Perrott, F.A., Perrott, R.A., Stokes, S., Waller, Heusser, L.E., SchluK chter, C., Marchant, D.R., Denton, G.H., 1995. M.P., Huang, Y., Eglington, G., Ivanovich, M., 1999. Environmental Interhemispheric correlation of Late Pleistocene Glacial events. change, lake and groundwater in the Sahel of Northern Nigeria. Science 269, 1541}1549. Part 2. Holocene landscape evolution of the Manga grasslands. Maley, J., 1991. The African rain forest vegetation and palaeoenviron- Evidence from palaeolimnology and dune chronology. Journal of ments during late quaternary. Climatic Change 19, 79}98. the Geological Society, London 155, 357}358. Maley, J., 1997. Middle to Late Holocene changes in tropical Africa and Holmgren, K., KarleH n, W., Shaw, P., 1995. Paleoclimatic signi"cance other continents: paleomonsoon and sea surface temperature vari- of the stable isotopic composition and petrology of a late Pleis- ations. In: Third Millenium BC Climate Change and Old World tocene stalagmite from Botswana. Quaternary Research 43 (3), Collapse. NATO ASI Series. Series 1, Vol. 49, pp. 611}640. 320}328. Maley, J., Brenac, P., 1998. Vegetation dynamics, palaeoenvironments Hulme, M., 1992. Rainfall changes in Africa: 1931}1960 to 1961}1990. and climatic changes in the forests of western Cameroon during the International Journal of Climatology 12, 685}699. last 28,000 years B.P. Review of Palaeobotany and Palynology 99, HsuK , C.-P.F., Walace, J.M., 1976. The global distribution of the annual 157}187. and semiannual cycles in precipitation. Monthly Weather Review Margraft, V., Dodson, J.R., Kershaw, A.P., Mcglone, M.S., Nicholls, N., 104/9, 1093}1101. 1991. Evolution of late Pleistocene and Holocene climates in Johnsen, S.J., Clausen, H.B., Dansgaard, W., Fuhrer, K., Gundestrup, the circum-South Paci"c land areas. Climate Dynamics 6, N., Hammer, C.U., Iversen, P., Jouzel, J., Stau!er, B., Ste!ensen, 193}211. 210 F. Gasse / Quaternary Science Reviews 19 (2000) 189}211

Martin, L., Fournier, M., Mourguiart, P., Sifeddine, A., Turcq, B., Ansy, ganique lacustre en zone tropicale sud au cours des 36.000 dernie`res M.L., Flexor, J.M., 1993. Southern Oscillation signal in South anneH es (Lac Tritrivakely, Madagascar). Comptes Rendus del' Aca- American palaeoclimatic data of the last 7000 years. Quaternary deH mie des Sciences, Paris 321, 385}391. Research 39, 338}346. Sirocko, F., Garbe-SchoK nberg, C.-D., McIntyre, A., Mol"no, B., 1996. Mayewski, P.A., Meeker, L.D., Whiltov, S., Twickler, M.S., Morrison, Teleconnections between the subtropical monsoons and high-latit- M.C., Bloom"eld, P., Bond, G.C., Alley, R.B., Gow, A.J., Grootes, ude climates during the last deglaciation. Nature 272, 526}529. P.M., Meese, D.A., Ram, M., Taylor, K.C., Wumkes, W., 1995. Servant, M., Servant-Vildary, S., 1980. L'environnement quaternaire du Changes in atmospheric circulation and ocean ice cover over the bassin du Tchad. In: Williams, M.A.J., Faure, H. (Eds.), The Sahara North Atlantic during the last 41,000 years. Science 263, 1747}1751. and the Nile. Balkema, Rotterdam, pp. 133}162. McIntyre, A., Ruddiman, W.F., Karlin, K., Mix, A.C., 1989. Surface Sonntag, C., Klitzsch, E., Lohnert, E.P., El Shazly, E.M., Munnich, K.O., water response of the equatorial Atlantic Ocean to orbital forcing. Junghans, Ch., Thorweihe, U., Weistro!er,K.,Swailem,F.M.,1978. Paleoceanography 4, 19}55. Paleoclimatic information from deuterium and oxygen-18 in carbon- Mourguiart, P., Correge, T., Wirrmann, D., Argollo, J., Montenegro, 14-dated north Saharian groundwaters. In: Isotope Hydrology, Vol. M.E., Pourchet, M., Carbonel, P., 1998. Holocene palacohydrology 978. International Atomic Energy Agency, Vienna, pp. 569}580. of Lake Titicaca estimated from an osttracod-based trasnfer func- Sowers, T., Bender, M., 1995. Climate records covering the Last De- tion. Palaeography, Palaeoclimatology, Palaeoecology 143, 51}72. glaciation. Science 269, 210}214. Nicholson, S.E., 1996. A review of climate dynamics and climate varia- Stager, J.C., Mayewski, P.A., 1997. Abrupt early and Mid-Holocene bility in Eastern Africa. In: Johnson, T.C., Odada, E.O. (Eds.), The climatic transition registered at the equator and the poles. Science Limnology, Climatology and Paleoclimatology of the East African 276, 1834}1836. Lakes. Gordon and Breach, Amsterdam, pp. 25}56. Steig, E.J., Brook, E.J., White, J.W.C., Sucher, C.M., Bender, M.L., Ning Shi, Dupont, L.M., Beug, H.-J., Schneider, R., 1998. Vegetation Lehman, S.J., Morse, D.L., Waddington, E.D., Clow, G.D., 1998. and climate changes during the last 21 000 years in S.W. Africa Synchronous climate changes in Antarctica and the North Atlantic. based on a marine pollen record. Veget. Hist. Archaeobot. 7, Science 282, 92}95. 127}140. Street-Perrott, F.A., Perrott, R.A., 1990. Abrupt climate #uctuations in O'Brien, P.A., Mayewski, P.A., Meeker, L.D., Meese, D.A., Twickler, the tropics: the in#uence of Atlantic Ocean Circulation. Nature 343, M.S., Whitlow, S.I., 1995. Complexity of Holocene climate as recon- 607}611. structed from a Greenland Ice core. Science 270, 1962}1964. Street-Perrott, F.A., Perrott, R.A., 1993. Holocene vegetation, lake Pachur, H.J., Altmann, N., 1997. The Quaternary (Holocene, ca. 80000 levels and climate in Africa. In: Global Climates Since the Last a B.P.). In: Schandelmeier, H., Reynolds, P.O. (Eds.), Palaeogeo- Glacial Maximum. University of Minnesota Press, Minneapolis, pp. graphic-Palaeotectonic Atlas of North-Eastern Africa, Arabia, and 318}356. adjacent areas. Balkema, Rotterdam, pp. 111}125. Street-Perrott, F.A., Marchand, D.S., Roberts, N., Harrison, S.P., 1989. PAGES, 1997. PEP III: The Pole-Equator-Pole transect through Euro- Global lake-level variations from 18,000 to 0 years ago: a Palaeocli- pe and Africa. In: Gasse, F., Kroepelin, S., Old"eld, F. (Eds.), matic analysis. In: United States Department of Energy (Eds.), PAGES Workshop Report. Series 97}2, Bern, pp. 64. Washington, USA, DOE/ER/60304-HI, pp. 1}213. Partridge, T.C., 1997. Cainozoic environmental change in southern Street-Perrott, F.A., Huang, Y., Perrott, R.A., Eglington, G., Barker, P., Africa, with special emphasis on the last 200 000 years. Progress in Ben Khelifa, L., Harkness, D.D., Olago, D.O., 1997. Impact of lower Physical Geography 21 (1), 3}22. atmospheric CO on tropical mountain ecosystems. Science 278, Partridge, T.C., deMenocal, P.B., Lorentz, S.A., Paiker, M.J., Vogel, 1422}1426. J.C., 1997. Orbital forcing of climate over South Africa: a 200 Stuiver, M., Reimer, P.J., 1993. Extended C data base and 000-year rainfall record from the Pretoria Saltpan. Quaternary revised CALIB 3.0 C age calibration program. Radiocarbon 35, Science Reviews 16, 1}9. 215}230. Pastouret, L., Chamley, H., Delibrias, G., Duplessy, J.C., Thiede, J., Stute, M., Talma, S., 1998. Glacial temperatures and moisture transport 1978. Late Quaternary climatic change in Western Tropical Africa regimes reconstructed from noble gas and dO, Stampriet aquifer, deduced from deep-sea sedimentation o! the Niger delta. Namibia. In: Isotope Techniques in the Study of Past and Current Oceanological Acta 1 (2), 217}232. Environmental Changes in the Hydrosphere and the Atmosphere. Petit-Maire, N., Page, N., Marchand, J., 1993. The Sahara in the Proceedings of Vienna Symposium 1997, IAEA, Vienna, SM- Holocene: Map 1/5,000,000, Marseille, Laboratoire de GeH ologie du 349/53, pp. 307}328. Quaternaire, CNRS. Sultan, M., Stucchio, N., Hassan, F.A., Hamdan, M.A., Mahmoud, Pichon, J.J., Labeyrie, L.D., Bareille, G., Labracherie, M., Duprat, J., A.N., Alfy, Z., Stein, T., 1997. Precipitation source inferred from Jouzel, J., 1992. Surface water temperature changes in the high stable isotopic composition of Pleistocene groundwater and car- latitudes of the Southern Hemisphere over the last glacial-inter- bonate deposition in the Western desert. Quaternary Research 48, glacial cycle. Paleoceanography 7, 289}318. 29}37. Roberts, N., Taieb, M., Barker, P., Damnati, B., Icole, M., Williamson, Sylvestre, F., Servant, M., Servant-Vildary, S., Causse, C., Fournier, M., D., 1993. Timing of the Younger Dryas event in East Africa from Ybert, J.P., 1998. Lake-level chronology on the southern Bolivian lake-level changes. Nature 366, 146}148. Altiplano (18}233S) during the Late Glacial and early Holocene. Schneider, R., MuK ller, P.J., Ruhland, G., 1995. Late Quaternary surface Quaternary Research 51, 54}66. circulation in the eastern equatorial South Atlantic: evidence from Talbot, M.R., Johannessen, T., 1992. A high resolution palaeoclimatic alkenone sea surface temperature. Paleoceanography 10, 197}219. record of the last 27,500 years in tropical West Africa from the Said, R., 1993. The River Nile. Pergamon, Oxford, pp. 320. carbon and nitrogen isotopic composition of lacustrine organic Scott, L., 1989. Climatic conditions in southern Africa since the last matter. Earth and Planetery Science Letters 110, 23}37. glacial maximum, inferred from pollen analysis. Palaeogeography, Talbot, M.R., Livingstone, D.A., 1989. Hydrogen index and carbon Palaeoclimatology, Palaeoecology 70, 345}353. isotopes of lacustrine organic matter as lake-level indicators. Scott, L., 1993. Palynological evidence for Late Quaternary warming Palaeogeography, Palaeoclimatology, Palaeoecology 80, 283}300. episodes in southern Africa. Palaeogeography, Palaeoclimatlogy, Talbot, M.R., Livingstone, D.A., Palmer, D.G., Maley, J., Melack, J.M., Palaeoecology 101, 229}236. Delibrias, G., Gulliksen, J., 1984. Preliminary results from sediments Sifeddine, A., Laggoun-DeH farge, F., Lallier-Verge`s, E., Disnar, J.R., core from lake Bosumtwi, Ghana. Palaeoecology of Africa 16, Williamson, D., Gasse, F., Gibert, E., 1995. La seH dimentation or- 173}192. F. Gasse / Quaternary Science Reviews 19 (2000) 189}211 211

Talma, A.S., Vogel, J.C., 1992. Late Quaternary paleotemperatures Valero-Garce`s, B.L., Grosjean, M., Schwalb, A., Geyh, M., Messerli, B., derived from a speleothem from Cango caves, Cape province, South Kelts, K., 1996. Limnogeology of Laguna Miscanti: evidence for Africa. Quaternary Research 37, 203}213. mid to late Holocene moisture changes in the Atacama Altiplano Texier, D., de Noblet, N., Harrison, S.P., Haxeltine, A., Jolly, D., Laarif, (Northern Chile). Journal of Paleolimnology 16, 1}21. F.I.C.P., Tarasov, P., Joussaume, S., 1997. Quantifying the role of Vincens, A., 1991. Late Quaternary vegetation history of the South- biosphere-atmosphere feedbacks climate change: coupled model Tanganyika basin. Climatic implications in South Central Africa. simulations for the 6000 years BP and comparison with paleodata Palaeogeography, Palaeoclimatology, Palaeoecology 86, 207}226. for North Eurasia and Northern Africa. Climate Dynamics 13, Wei, K., Gasse, F., 1999. Oxygen isotopes in lacustrine carbonates of 865}875. West China revisited. Implications for post glacial changes in summer Thompson, L.G., Mosley-Thompson, E., Davis, M.E., Lin, P.-N., monsoon circulation. Quaternary Science Review 18, 1315}1334. Henderson, K.A., Vole-Dai, J., Bolzan, J.F., Liu, K.-b., 1995. Late Williams, M.A.J., Adamson, D.A., Williams, F.M., Morton, W.H., Parry, Glacial stage and Holocene tropical ice core records from Huas- D.E., 1980. Jebel Marra Volcano: a link between the Nile Valley, the caran, Peru. Science 269, 46}50. Sahara and Central Africa. In: Williams, M.A.J., Faure, H. (Eds.), The Tyson, P.D., 1991. Climatic change in southern Africa: past and present Sahara and the Nile. Balkema, Rotterdam, pp. 305}337. conditions and possible future scenarios. Climate Change 18, Williamson, D., Jelinowska, A., Kissel, C., Tucholka, P., Gibert, E., 241}258. Gasse, F., Massault, M., Taieb, M., Van Campo, E., Wieckowski, K., Tyson, P.D., Gasse, F., Bergonzini, L., d'Abreton, P., 1997. Aerosols, 1998. Rock magnetic proxies of erosion/oxidation cyles in Late atmospheric transmissivity and hydrological modelling of climate Quaternary maar lake sediments (Lake Tritrivakely Madagascar) change over Africa South of the Equator. International Journal of paleoenvironmental implications. Earth and Planetary Science Let- Climatology 17, 1651}1665. ters 155, 205}219.